Stable diffusion

What is Stable Diffusion Online?

Stable Diffusion Online is a web-based AI platform for creating images and videos, operated by Black Technology LTD. Users can enter prompts or upload reference images to generate pictures, variations of those images, and short videos; they can also use editing tools such as zooming, enlarging images, and removing backgrounds.

The official website listed in the database is not the official site of Stability AI, the developer of the Stable Diffusion model. The web platforms, the various WebUIs within the Stable Diffusion model family, and the community are distinct entities; therefore, the licensing terms, privacy policies, and billing entities cannot be conflated.

A one-sentence summary

Text-to-image, image-to-image, AI video generation, and various common image processing tasks can be carried out in the browser without the need for a local graphics card; additional capabilities and more models are available through point systems and membership plans.

Core functions

• Generate illustrations, photographs, posters, and concept art based on textual descriptions.
• Upload a reference image and create a new version that retains the main elements or style.
• Choose styles such as Automatic, Design, Cinematic, Realistic, Animation, and Comic.
• Switch between various horizontal, vertical, and square aspect ratios.
• Multiple candidate results are generated at once, which facilitates the comparison of compositions.
• Generate AI short videos from text or reference frames.
• Use an image editor to further adjust the existing materials.
• Common tasks are carried out through background removal, image enlargement, and scaling.
• Search large prompt libraries and refer to community works.
• Connect static diagrams, editing functions, and dynamic processes within the same workspace.

Text-to-image feature

Text-to-image generation is suitable for those who want to explore visual concepts from scratch. By entering details such as the subject, environment, lighting, camera angle, materials, and style, the system generates candidate images based on the selected model, proportions, and number of outputs.

• It supports the entry of prompt text of up to about 500 characters.
• Different generation models and preset styles can be selected.
• Common aspect ratios such as 1:1, 4:3, 3:4, 16:9, and 9:16 are available.
• Supports output of one or multiple results.
• Suitable for profile pictures, product displays, covers, posters, and social media images.
• Generating tasks consumes points based on the model and settings.

Image-to-image and reference image

Image-to-image generation uses existing images as a structural or visual reference, making it suitable for preserving the posture of people, the shape of products, the layout, or the overall style. It does not guarantee a pixel-for-pixel replication; the identity of individuals, text, and details may still change.

• Add materials, lighting, and finishing touches to the sketch.
• Convert the same composition into different artistic styles.
• Create marketing scenarios around the product reference images.
• Explore variations in clothing, background, and atmosphere for character photos.
• Continue to create variants based on the results of the previous version.
• Try to preserve the original image structure by minimizing the extent of changes.

AI video and dynamic creation

The platform offers tools for generating video from text and from images; it enables the conversion of static concepts into short videos, while also allowing for the analysis of camera movement, rhythm, and composition. The terms of use apply, and the video generation function may make use of models such as Veo 3, SeedDance, and Sora.

• Generate short-video drafts from text scenarios.
• Upload the first frame or a reference image to create dynamic effects.
• Create product displays, social media videos, and ad concepts.
• Test different motion schemes for the same image.
• Feed the generated images into the video processing pipeline.
• The integral consumption of advanced models is usually higher than that of basic image generation.

Image editing tools

How to create the first image

1. Enter the image generation workspace of Stable Diffusion Online.
2. Log in to your account and check your current available points.
3. Choose to start from text, or upload a reference image for which you have usage rights.
4. Express the subject, action, environment, and target style in one sentence.
5. Select the model, style, aspect ratio, and number of outputs.
6. Check the required points for the task and submit it to generate it.
7. Select from the candidate images the one whose composition is closest to the target.
8. Modify the prompt or add reference images to continue iterating.
9. Use zoom, background removal, or image enlargement as needed to finish the task.
10. Check for characters, text, trademarks, and visual defects before downloading.

Prompt writing format

High-quality prompt texts do not need to be filled with adjectives; the focus should be on clearly specifying the visual information that needs to be retained. It is recommended to organize such prompts by subject, action, scene, composition, lighting, material, style, and purpose.

Prompt iteration method

1. First, use brief prompts to confirm the subject and the composition direction.
2. Change only one variable at a time, such as the light or the lens.
3. Keep the relevant keywords and remove descriptions that have no clear purpose.
4. Use reference diagrams to illustrate structures that are difficult to describe in words.
5. Leave clear space and define the position of the main elements for posters and product images.
6. Perform local inspections on faces, hands, text, and brand elements.
7. Record the model, scale, prompts, and generation results.
8. Select the direction first, then zoom in and make fine adjustments to avoid wasting points prematurely.

Prompt database

The website offers a large database of prompt templates; the current page claims to contain over ten million visual generation prompts. It is useful for finding ideas regarding composition and style, but one must assess on their own the accuracy of the community-generated content, its licensing status, and its compatibility with various models.

• Search for images and prompts by theme.
• Learn about photography, illustration, and key phrase combinations for a cinematic style.
• Break down the reference prompts and use them in your own creations.
• Do not copy directly content that includes names of persons, trademarks, or claims of infringement.
• Different models may respond very differently to the same prompt.
• Community-created content does not mean that the platform guarantees anything regarding its copyright status.

Supported model and brand boundaries

The website’s name emphasizes Stable Diffusion, but the services offered currently do not rely solely on it. The official policies allow for other technologies to be used; for image generation, Nano Banana and SeedDream series may be employed, while for video generation, Veo 3, SeedDance and Sora series can be used.

Free version and membership prices

The following shows the annual billing amounts as verified in real time on the official pricing page in August 2026, with values expressed in US dollars. The page indicates that an annual payment can result in a savings of around 50%; the price for monthly payments is not consistently displayed in the data captured, so no estimate can be provided for it.

Public rights and interests of various plans

The pricing page currently presents three options: ad-free, watermark-free versions along with advanced models, commercial licenses, and private generation; the main differences lie in the number of points awarded and the quantity of outputs per transaction. It is still advisable to review the benefits on the payment page before making a purchase, as the details listed there may change.

How to spend points

The amount of points consumed depends on the model, resolution, quantity, and type of task. When checking the text-to-image generation page, the current example shows that one task requires 2 points; tasks involving videos, advanced models, multiple results, or high resolution may require more points.

• Check the points required for this task as displayed on the interface before submitting.
• The free version provides 10 points per day, which may not be sufficient to cover tasks that involve high-cost videos.
• First, use a low-cost setup to verify the direction of the prompt.
• Generating multiple cards at once increases the number of candidates and accelerates the consumption of points.
• It is the account records and the instructions provided by customer service that determine whether points are refunded for failed tasks.
• It is the settlement page that determines whether member points are carried over and when they are updated.

Subscriptions, refunds, and account rules

The terms of service state that paid plans, points, and subscriptions are charged at the prices displayed on the checkout page. Unless required by law, purchases are generally non-refundable; furthermore, the platform may suspend or terminate accounts due to security, legal, abuse, payment, or policy-related risks.

• Before purchasing, confirm whether it is a monthly or annual payment plan and the total amount that will be deducted.
• Save the order details, plan benefits, point count, and renewal date.
• Do not give your account or payment details to others for use.
• Confirm when the benefits will cease before canceling the subscription.
• Purchased points and subscriptions may have different expiration rules.
• A refund is not an automatic right; in case of disputes, contact support as soon as possible.
• Violating the content policy may result in rejected submissions or account termination.

Commercial use and copyright

The pricing page lists commercial licensing as one of the benefits of each plan, but users still need to ensure that they possess the necessary rights to the prompts, uploaded materials, portraits, trademarks, and reference images. The fact that the platform allows commercial use does not mean that the generated content will automatically be protected by copyright, nor does it eliminate the risk of infringement by third parties.

• Do not upload customer materials or photographic works without authorization.
• Obtain permission for portraits and privacy when dealing with real people.
• Avoid creating logos that could be confused with existing brands.
• Check for any unintended text or trademarks in the images before publishing.
• Save prompts, models, dates, and records of manual modifications for business projects.
• Different countries have varying regulations regarding the copyright of content generated purely by AI.
• High-risk advertisements, news, and political content require additional legal review.

Stability AI model license

If users download and deploy Stability AI’s Core Models on their own, the Stability AI license applies instead of the terms of service for Stable Diffusion Online. The current Community License allows individuals or organizations with an annual income of less than $1 million to use these models for research, as well as for non-commercial and commercial purposes, free of charge.

Stability AI states that, in the relationship between users and Stability AI, the outputs generated using Core Models or their derivatives in a compliant manner belong to the users. However, this provision in the contract does not necessarily mean that copyright authorities in various regions will recognize such AI-generated outputs as having copyright protection.

Privacy and content moderation

The generation page claims that the anonymous mode does not collect or store any text or images, but the terms of service allow the platform to monitor usage in order to enforce policies; the acceptable use policy also specifies that automated systems as well as manual review can be used to examine prompts, uploaded content, and the generated results.

• Do not include trade secrets, identification documents, or personal sensitive information in the prompts.
• Remove the location data and device metadata from the photos before uploading.
• Do not assume that all functions have zero retention based solely on the anonymous statements on the marketing page.
• Paid accounts still generate the necessary account and transaction records.
• Policy review may check the prompts, uploaded materials, and generated results.
• For confidential projects, it is necessary to first confirm with the authorities the mechanisms for retention, training, and deletion.
• Teams with high privacy requirements can consider deploying compliant local models.

Content security rules

• Pornography, nudity, fetishism, adult content, and explicit material are prohibited.
• Sexual content involving minors or individuals of unknown age is strictly prohibited.
• Content involving non-consensual sexualization, sexual violence, and exploitation is prohibited.
• Hate, harassment, discrimination, and violent or bloody content are prohibited.
• Self-harm, illegal activities, fraud, and misleading content are prohibited.
• One must not impersonate others or violate privacy and intellectual property rights.
• It is not allowed to bypass the platform’s security restrictions and review mechanisms.
• Violating requests may be blocked, the content may be removed, and the account may be suspended.

What use cases are suitable?

• Content creators create article covers and images for social media.
• The designer quickly explores various directions for posters and visual styles.
• The e-commerce team creates product scenarios and advertising concepts.
• The video team converted the static storyboards into dynamic tests.
• Marketing staff create multiple versions of draft visual materials for the campaign.
• Illustration enthusiasts try different styles and compositions.
• Teachers create schematic diagrams for the courses that contain no sensitive information.
• Users without local high-performance graphics cards cannot use the online experience generation model.
• Beginners learn visual description methods through a library of prompt examples.

Tasks that are not suitable for completion directly

• A final design draft that requires fully controllable pixels and precise text formatting.
• Batch materials requiring 100% consistency in character identities and product structures.
• Enterprise projects that contain confidential data and do not have a data processing agreement in place.
• Scenarios that require offline operation, private deployment, or full control over the source code.
• Adults, pornography, blood, hate, or other prohibited content.
• Projects that replicate living artists without authorization or use protected characters.
• News and evidence images that require authoritative accuracy.
• Research tasks that require a fixed model version and reproducible experimental parameters.

Product advantages

• It can be used directly in a browser; there is no need to install models or configure graphics cards.
• Image generation, image conversion, video, and editing functions are all available on one platform.
• A variety of styles, proportions, and output quantities are available for selection.
• It includes practical tools such as background removal, zooming, and image enlargement.
• The free version provides a small amount of points per day to facilitate testing.
• A large library of prompt examples helps beginners learn.
• The difference in points between the annual Pro and Max plans is quite clear.
• The current website offers a Chinese interface.
• There is no need to manage local model files or the runtime environment.

Usage restrictions and precautions

• The website is not an official service of Stability AI, and its brand name may cause confusion.
• The current platform may invoke multiple third-party and internal integration models.
• The integration, mass, and availability of different models may vary.
• AI-generated faces, hands, text, and complex structures may contain errors.
• The reference images do not guarantee that the people or products will be exactly identical.
• The monthly price is not displayed consistently in the results of this extraction.
• Unless required by law, purchases are generally non-refundable.
• The privacy statements on marketing pages and the review policies should be read together.
• A commercial license does not automatically resolve issues related to third-party copyright and portrait rights.
• The service is provided as it is, without any guarantee of continuous availability or error-free operation.
• The platform does not make the complete source code available, which makes it impossible to conduct an audit of its online implementation on one’s own.

GitHub and the open-source status

On the Stable Diffusion Online website, no complete source code repository maintained by Black Technology LTD was found; therefore, this site should be classified as non-open source. The fact that the website can use Stable Diffusion or other models does not mean that its code is open source.

Stability AI maintains multiple official model and reference implementation repositories. For example, the official repository for SD3.5 provides reference code for inference under the MIT license, but the model weights are governed by the Stability AI Community License; it is necessary to check both the open-source nature of the code and the licensing terms applicable to the weights separately.

How to choose between local deployment and the online version?

Basic information

Recommendation score

4.2 / 5. Stable Diffusion Online is suitable for users who do not want to configure a graphics card and wish to quickly create drafts of images and videos; it offers a fairly comprehensive set of functions. However, it is not the official website of Stability AI, so details such as the source of the models, the cost associated with points, privacy policies, and commercial licensing terms need to be checked for each specific task.

Frequently Asked Questions

Is Stable Diffusion Online free?

A free plan is available; 10 points are awarded per day for verification, and 2 images can be generated at a time. Different tasks require different amounts of resources, and the free quota may be exhausted quickly when dealing with high-quality images or videos.

How much are the Pro and Max versions?

When checking the annual payment page in August 2026, Pro will cost $10 per month and include 2,000 points per month, while Max will cost $20 per month and include 5,000 points per month. The actual total amount deducted and the monthly price will be as indicated on the settlement page.

Is it the official website of Stability AI?

No. Stable Diffusion Online is operated by Black Technology LTD, while Stability AI is the official entity responsible for the development and licensing of the Stable Diffusion model family.

Does the platform use only Stable Diffusion?

No. According to the current policy, the image and video features may integrate models from series such as Nano Banana, SeedDream, Veo 3, SeedDance, and Sora.

Can the generated images be used for commercial purposes?

The pricing page indicates a commercial license, but users remain responsible for the materials they upload, as well as for any portraits, trademarks, and the manner in which content is published. Specific cases also require consideration of local copyright rules and the rights of third parties.

Does the website save the prompts and images?

The generation page claims to be anonymous and does not store any text or images, but the policies allow for the inspection of prompts, uploaded content, and results for security purposes. In cases involving confidential material, it is necessary to consult the platform first to determine the specific retention procedures.

Does it support NSFW content?

Not supported. Adults, pornography, nudity, sexual suggestions, and explicit content are all prohibited; requests that violate these rules may be blocked or result in the suspension of the account.

Is Stable Diffusion open source?

It is necessary to distinguish between code and weight licenses. Some official reference implementation codes are made available publicly, but the model weights are subject to their respective licenses; the Stable Diffusion Online website itself does not provide the complete source code.